Lung Function Impairment and Risks of Incident Cardiovascular Diseases and Mortality Among People With Type 2 Diabetes: A Prospective Cohort Study Free

Type 2 diabetes (T2D) is a chronic metabolic disorder with a rapidly increasing global prevalence. As of 2021, it has affected ∼537 million adults worldwide, a number projected to rise to 783 million by 2045 (1). This alarming increase highlights the growing public health burden posed by T2D and its associated complications, including cardiovascular disease (CVD), which remains the leading cause of morbidity and mortality in this population (2). While complications such as nephropathy, retinopathy, and neuropathy are well-recognized, emerging evidence has suggested that the lungs may also be a target organ affected by diabetes (3,4). Mechanisms such as microvascular damage, chronic inflammation, and glycosylation of lung tissues may contribute to impaired lung function, increasing the risk of restrictive and obstructive lung diseases in individuals with T2D (5). Reduced lung function, typically measured by forced expiratory volume in 1 s (FEV₁) and forced vital capacity (FVC), is associated with increased CVD risk and mortality (6–9). However, how this impairment specifically affects cardiovascular outcomes in people with T2D remains insufficiently explored.

Although the relationship between lung function impairment and cardiovascular outcomes has been extensively studied in the general population (10–14), research focusing on individuals with T2D is limited. Early research by Klein et al. (15) and Davis et al. (16) offered valuable insights into the association between pulmonary function and diabetes-related outcomes. However, these studies were limited by small sample sizes, relatively short follow-up durations, and a primary focus on all-cause mortality or general pulmonary measures rather than cardiovascular outcomes. While a few more recent studies have attempted to address these gaps, they are also frequently constrained by methodological limitations, such as cross-sectional designs, which limit the ability to draw definitive conclusions (17,18). Moreover, there is a lack of longitudinal data on how different forms of lung function impairment influence long-term cardiovascular outcomes and mortality in this high-risk group (19). The aim of this prospective cohort was to determine whether lung function impairment is an independent predictor of CVD and mortality in individuals with T2D using data from the UK Biobank. We also examined whether incorporating lung function into the Systematic Coronary Risk Evaluation 2 (SCORE2)–Diabetes model, which is specifically tailored for European populations with diabetes (20), can improve its ability to predict CVD risk.

The UK Biobank is a large, prospective cohort study that has successfully recruited more than a half a million middle-aged and older adults from 22 centers across England, Scotland, and Wales between 2006 and 2010 (21). It has systematically gathered a diverse array of information, including sociodemographic characteristics, physical examinations, medical history, biological samples, and imaging data. Data were collected using various methods, such as digital and touch-screen questionnaires, oral interviews, clinical records, and physical measurements. Ethical approval for the UK Biobank was obtained from the North West Multicenter Research Ethics Committee (reference number 11/NW/0382), with all participants giving written informed consent. The current study was performed under application number 150464. In the current analysis, we examined data of 502,244 participants in the UK Biobank and excluded those without written consent and valid spirometry data. Participants who were not classified as having T2D or had CVD at baseline were also excluded, leaving a total of 16,242 participants for analysis (Supplementary Fig. 1). The descriptions, definitions, and field identifiers (unique numeric codes for specific variables in the UK Biobank) for defining T2D and CVD at baseline in the UK Biobank are presented in Supplementary Table 1.

Lung function was assessed using a Vitalograph Pneumotrac 6800 spirometer, adhering to standardized protocols under the supervision of trained health care professionals. Participants were instructed to perform two to three forceful expiratory blows within a 6-min allocated window. The reproducibility of the first two blows was evaluated, and if the difference between them was <5%, a third blow was not required. To ensure accuracy, blows were considered valid only if they met specific criteria: the extrapolated volume at the start was not excessive, the time to peak flow was appropriate, a sufficient plateau at the end of the test was achieved, and no cough was detected during the maneuver. Additionally, blows <6 s or those explicitly rejected by the investigator were deemed invalid. From the valid blows, the highest values for FEV₁ and FVC, referred to as the best measure, were used for analysis to ensure the most accurate reflection of lung function (11).

Using the Global Lung Function Initiative 2012 reference equations, which account for age, sex, height, and ethnicity (22), we calculated the predicted percentages for FEV₁ (FEV₁% predicted) and FVC (FVC% predicted), along with their respective lower limit of normal (LLN) thresholds. Lung function impairment was then classified into clinically meaningful groups (lung function category 1), including preserved lung function (defined as FEV₁/FVC ≥ LLN and FVC ≥ LLN), obstructive physiology (OP) (defined as FEV₁/FVC < LLN), and restrictive physiology (RP) (defined as FEV₁/FVC ≥ LLN and FVC < LLN) (11). In addition, we defined preserved ratio impaired spirometry (PRISm) (FEV₁/FVC ≥ LLN and FEV₁ < LLN) and its control counterparts (FEV₁/FVC ≥ LLN and FEV₁ ≥ LLN) (23), which we classified as lung function category 2.

The primary outcomes of this study were incident CVD and all-cause mortality. Incident CVD was defined as a composite end point consisting of nonfatal myocardial infarction (MI), stroke, heart failure (HF), and cardiovascular death. Secondary outcomes were the individual components of CVD, including nonfatal MI, nonfatal stroke, nonfatal HF, and cardiovascular death. The ascertainment of CVD events was extracted from hospital episode records and death registers, determined using ICD-9 and ICD-10 codes. Diagnostic and procedural codes for primary and secondary outcomes are shown in Supplementary Table 2. The follow-up period was calculated from the enrollment date (baseline) to the occurrence of incident CVD, all-cause death, or the end of follow-up on 30 June 2023, whichever occurred first.

The study included a variety of covariates, including age (years), sex (female or male), marital status (married or not), Townsend deprivation index (TDI), education (college and above or not), smoking status (current, former, or never), drinking status (current, former, or never), regular physical activity (yes or no), healthy diet score (0–1, 2–3, and 4–5), sleep duration (hours), chronic lung disease (including chronic obstructive pulmonary disease, asthma, chronic bronchitis, and emphysema) (11), sleep apnea, systolic blood pressure (SBP) (mmHg), diastolic blood pressure (mmHg), total cholesterol (TC) (mmol/L), HDL cholesterol (HDL-C) (mmol/L), glycated hemoglobin (HbA_1c) (mmol/mmol), age at diabetes diagnosis (years), BMI (kg/m²), estimated glomerular filtration rate (eGFR) (mL/min/1.73 m²), C-reactive protein (CRP) (mg/L), hypertension, diabetes medication (classified into no drug use, metformin alone, insulin alone, sulfonylureas alone, other monotherapy, and dual therapy or higher), antihypertensive medication (grouped into no drug use, ACE inhibitors alone, angiotensin receptor blockers alone, calcium channel blockers alone, β-blockers alone, other monotherapy, dual therapy, and triple therapy or higher), cholesterol-lowering medication (divided into no drug use, statins, and other), aspirin use, and respiratory medication (including bronchodilators, oral corticosteroids, and inhaled corticosteroids) (24).

Missing covariates were imputed using multiple imputation by chained equations (see Supplementary Methods), and the numbers (percentages) of missing covariates are displayed in Supplementary Table 3. Descriptive characteristics of the study participants are presented as median (interquartile range [IQR]) or number (percent) as appropriate. Baseline characteristics among study participants grouped by lung function category 1 were compared using χ² tests for categorical variables and the Kruskal-Wallis test for continuous variables.

We reported the crude incidence rates (95% CIs) of CVD and all-cause mortality as the number of events per 10,000 person-years, with time since baseline measurement as the timescale. The cumulative incidence of CVD across lung function impairment categories was calculated using the Fine-Gray proportional subdistribution hazards model, and the cumulative hazard of all-cause mortality was calculated using the Kaplan-Meier method. To assess the dose-response relationship and address potential nonlinearity between continuous spirometry parameters and outcomes, we applied restricted cubic splines based on Cox proportional hazard models, with three knots positioned at the 5th, 50th, and 95th percentiles.

Proportional hazards assumption was examined using Schoenfeld residuals, and no evidence of violation was observed. The associations of impaired lung function metrics with incident CVD were assessed using Fine-Gray proportional subdistribution hazards models accounting for the competing risk of noncardiovascular deaths, while Cox proportional hazards models were used to evaluate the associations with all-cause mortality. Each lung function variable was added to the models separately to evaluate its independent association. We fitted three models incorporating covariates under the guidance of the directed acyclic graph (Supplementary Fig. 2), where covariates were categorized based on whether they were considered potential confounders or mediators in the causal pathway between lung function impairment and outcomes (25). Model 1 was a basic model that included age and sex. Model 2 included age, sex, marital status, TDI, education, smoking, drinking, physical activity, healthy diet score, sleep duration, chronic lung diseases, and sleep apnea. Model 3 additionally adjusted for covariates that were considered potential mediators, including SBP, diastolic blood pressure, TC, HDL-C, HbA_1c, age at diabetes diagnosis, BMI, eGFR, CRP, hypertension, diabetes medication, antihypertensive medication, cholesterol-lowering medication, aspirin, and respiratory medication. To avoid potential overadjustment for variables that could mediate associations between lung function impairment and outcomes, models 1 and 2 were conducted as the primary analysis and model 3 as the sensitivity analysis (25). Additionally, Harrell C-statistic, continuous net reclassification improvement (NRI), and absolute integrated discrimination improvement (IDI) were used to assess whether adding the lung function parameters could improve risk discrimination and reclassification for CVD prediction above the SCORE2-Diabetes model (26).

Finally, subgroup analyses and a series of sensitivity analyses were conducted. First, numeric cutoffs were applied to define various lung function categories, including preserved lung function (FEV₁/FVC ≥0.70 and FVC% predicted ≥80%), OP (FEV₁/FVC <0.70), RP (FEV₁/FVC ≥0.70 and FVC% predicted <80%), and PRISm (FEV₁/FVC ≥0.70 and FEV₁% predicted <80%) and its control group (FEV₁/FVC ≥0.70 and FEV₁% predicted ≥80%) (13). Second, participants who had incident CVD or died within the initial 2 years of follow-up were excluded to reduce the potential for reverse causality. Third, participants with chronic lung diseases or sleep apnea or who were taking respiratory medication were excluded from analyses to mitigate the influence of these risk factors on outcomes. Fourth, study centers were additionally adjusted and treated as a random intercept using Cox frailty models to account for clustering effects. Fifth, handling missing covariates as a separate category for categorical variables and imputed with mean values for continuous variables and restricting analyses to participants with nonmissing covariates were used as different methods for handling missing covariates. Sixth, covariates from model 3 were further adjusted to assess the robustness of our results. Finally, the QResearch Cardiovascular Risk Prediction Algorithm 3 (QRISK3) model (27) was used to evaluate the improvement in CVD risk prediction performance, given its robust validation in U.K. populations and inclusion of additional variables not captured by SCORE2-Diabetes, offering a broader comparison of risk prediction models.

All analyses were performed using R 4.4.0. A two-sided P < 0.0125 (0.05/4) for multiple testing correction using the Bonferroni method was considered statistically significant for the analyses of secondary outcomes. For all other analyses, a two-sided P < 0.05 was considered statistically significant.

Of 16,242 individuals with T2D (median [IQR] age 61 [55–65] years; female 40.1%), 12,440 (76.6%) had preserved lung function, 1,377 (8.5%) had OP, and 2,425 (14.9%) had RP. Participants with impaired lung function tended to be male and not married, had lower education, were more likely to be current smokers, had higher CRP values, and were more likely to have prevalent chronic lung diseases and sleep apnea than those with preserved lung function (Supplementary Table 4).

During a median follow-up of 13.9 years, there were 2,825 (17.4%) cases of incident CVD and 2,811 (17.3%) all-cause deaths. The survival curves for CVD and all-cause mortality across lung function categories are displayed in Fig. 1. The incidence rates of CVD were 120.85 (95% CI 115.52–126.35), 185.26 (95% CI 165.24–207.03), and 196.97 (95% CI 181.26–213.67) per 10,000 person-years in those with preserved lung function, OP, and RP, respectively, with the corresponding values for control and PRISm groups being 119.43 (95% CI 114.07–124.98) and 194.38 (95% CI 179.73–209.90), respectively (Table 1). In model 2, each SD decrement of FEV₁, FVC, FEV₁/FVC ratio, FEV₁% predicted, and FVC% predicted was related to a 26%, 30%, 5%, 18%, and 21% increased risk of CVD, and a 32%, 32%, 8%, 24%, and 24% increased risk of all-cause mortality, respectively. Compared with preserved lung function, the adjusted hazard ratios (HRs) for CVD and all-cause mortality were 1.19 (95% CI 1.05–1.35) and 1.41 (95% CI 1.24–1.58) for OP and 1.47 (95% CI 1.33–1.65) and 1.57 (95% CI 1.43–1.73) for RP, respectively. Compared with the control group, the adjusted HRs for CVD and all-cause mortality were 1.20 (95% CI 1.06–1.36) and 1.42 (95% CI 1.26–1.60) for OP and 1.43 (95% CI 1.29–1.59) and 1.52 (95% CI 1.39–1.66) for PRISm, respectively (Table 1). The restrictive cubic splines showed negative relationships (linear or L-shaped) of lung function measures with CVD and all-cause mortality (Fig. 2). The associations between lung function impairment and individual outcomes, including MI, stroke, HF, and cardiovascular death, are presented at Supplementary Table 5.

The performance of all predictive models at 10 years after baseline are shown in Table 2. Adding FEV₁, FVC, FEV₁% predicted, and FVC% predicted to the SCORE2-Diabetes model significantly improved CVD risk prediction. The C-statistic was improved from 0.672 to 0.687, 0.687, 0.686, and 0.686, respectively. Furthermore, the continuous NRI was 0.207 (95% CI 0.144–0.261) for FEV₁, 0.195 (95% CI 0.123–0.256) for FVC, 0.045 (95% CI 0.012–0.117) for FEV₁/FVC, 0.150 (95% CI 0.087–0.202) for FEV₁% predicted, 0.161 (95% CI 0.105–0.207) for FVC% predicted, 0.162 (95% CI 0.105–0.225) for lung function category 1, and 0.181 (95% CI 0.102–0.233) for lung function category 2. Meanwhile, the IDI was 0.006 (95% CI 0.004–0.009) for FEV₁, 0.007 (95% CI 0.003–0.009) for FVC, 0.002 (95% CI 0.001–0.003) for FEV₁/FVC, 0.005 (95% CI 0.003–0.007) for FEV₁% predicted, 0.005 (95% CI 0.002–0.008) for FVC% predicted, 0.003 (95% CI 0.001–0.005) for lung function category 1, and 0.003 (95% CI 0.001–0.006) for lung function category 2 (Table 2).

In the stratified analyses, the associations of lung function measures and categories with risk of CVD and all-cause mortality remained consistent across most subgroups of selected covariates (Supplementary Tables 6 and 7). The main results did not change materially across a series of sensitivity analyses, which included using numeric cutoffs for defining lung function impairment categories (Supplementary Table 8); excluding participants who experienced outcomes within the initial 2 years of follow-up (Supplementary Table 9); excluding participants with chronic lung diseases, sleep apnea, or taking respiratory medication (Supplementary Table 10); addressing the clustering effects of study centers (Supplementary Table 11); using additional methods for handling missing covariates (Supplementary Table 12); adjusting for covariates from model 3 (Supplementary Table 13); and using the QRISK3 for evaluation of CVD risk prediction improvement (Supplementary Table 14).

In this prospective cohort study, we provide a comprehensive examination of lung function measures and impairment with risk of CVD and all-cause mortality in individuals with T2D. Our results show that lower lung function measurements were associated with an increased risk of CVD and all-cause mortality. In addition, OP, RP, and PRISm were all found to be associated with a higher risk of CVD and all-cause mortality in T2D compared with normal lung function. These associations were independent of traditional risk factors, such as smoking, and were consistent across subgroups and sensitivity analyses. Finally, our data show that adding the lung function measurements significantly enhanced the performance of CVD prediction beyond the SCORE2-Diabetes and the QRISK3 models.

Our findings are consistent with previous studies conducted in the general population, which have established a strong association between impaired lung function and increased CVD risk (7,11,13). Large-scale cohort studies have shown that reduced FEV₁ and FVC are significant predictors of CVD and mortality, independent of traditional risk factors such as smoking, blood pressure, and obesity (9,14). These studies suggested that diminished pulmonary function may contribute to systemic inflammation and oxidative stress, key mechanisms in the development of CVD. Our study extends these observations to individuals with T2D, demonstrating that impaired lung function in this population not only exacerbates cardiovascular risk but also increases the likelihood of all-cause mortality. This underscores the importance of considering lung function as a critical factor in CVD risk stratification, particularly in individuals with T2D who already face a heightened risk.

Compared with prior research focused on diabetes (17,18), our findings align with earlier studies that emphasized a strong link between reduced lung function and increased mortality and CVD risk. Notably, previous analyses have demonstrated that individuals with diabetes who had lower FEV₁ or FVC faced significantly higher risks of fatal CVD and all-cause mortality (18). However, these studies were based on cross-sectional designs, limiting their ability to capture the temporal relationship between lung function decline and cardiovascular outcomes. In contrast, our study, using a prospective cohort design, offers a more robust analysis of the long-term impact of lung function impairment, providing stronger evidence for the role of lung function as a predictor of adverse outcomes in T2D. Furthermore, our results align with findings from a recent study (19) on PRISm, which demonstrated significantly higher risks of macrovascular and microvascular complications, as well as cardiovascular and all-cause mortality, in individuals with T2D. The higher HRs for cardiovascular mortality and all-cause mortality in those with PRISm reinforce the critical role of lung function assessment, even in the absence of obstructive lung disease, for predicting cardiovascular outcomes in this population.

The link between reduced lung function and an elevated risk of CVD and mortality in patients with T2D can be attributed to several interrelated mechanisms. One of the primary pathways is systemic inflammation, which is often associated with impaired lung capacity (28,29). This chronic inflammation not only contributes to vascular dysfunction by damaging the endothelium and reducing arterial flexibility but also exacerbates insulin resistance and glucose metabolism disturbances in T2D, thus heightening cardiovascular risk (30–32). Lung function impairment can increase blood viscosity through elevated hematocrit levels, further promoting atherosclerosis and thrombosis and making the body susceptible to stroke and heart diseases (33,34). Additionally, shared risk factors, such as environmental exposures and smoking, play essential roles in both respiratory and cardiovascular health, though these associations persist even in nonsmokers, suggesting additional underlying biological processes (35,36). Moreover, restricted lung function can place additional strain on the heart by altering hemodynamic function, potentially leading to HF (37). These mechanisms collectively underscore the significant role that lung function plays in predicting cardiovascular complications and mortality among individuals with T2D.

Our study highlights the significant role of lung function impairment in CVD prevention among individuals with T2D. On one hand, our findings identify lung function impairment as an emerging cardiovascular risk factor in individuals with T2D, highlighting the potential of spirometry as a noninvasive and widely accessible tool for early CVD risk identification. On the other hand, our results align with existing guideline recommendations to address shared risk factors between pulmonary dysfunction and CVD, such as smoking and insufficient physical activity. Notably, PRISm remains an underrecognized condition, and our study highlights its significant association with CVD, underscoring the urgent need for early detection and integrated strategies to address both pulmonary and cardiovascular risks associated with PRISm. However, given that our analysis is based on a single baseline measure of lung function, the findings should be interpreted with caution and considered exploratory. Future longitudinal studies are necessary to capture dynamic changes in lung function over time and to establish their causal relationship with cardiovascular outcomes, ultimately providing stronger evidence for the integration of lung function measures into clinical risk management frameworks.

Our study has several important strengths. First, the use of data from a large, prospective cohort provided a robust foundation for analyzing the relationship between lung function impairment and cardiovascular outcomes in T2D. The large sample size and extensive phenotype data supported stratified analyses of different manifestations of lung function impairment, as well as subtypes of CVD. Additionally, the long follow-up period of up to 14 years enhanced the credibility of the findings. We also conducted multiple sensitivity analyses, and the results remained consistent across different scenarios, further supporting the robustness of our conclusions.

However, several limitations should be acknowledged. First, lung function was measured only once at baseline, restricting our ability to capture longitudinal changes or identify distinct lung function trajectories. This limitation may prevent us from fully understanding the dynamic relationship between pulmonary function decline and cardiovascular risk over time. Second, the use of administrative ICD-9 and ICD-10 codes for CVD outcome ascertainment may have led to misclassification due to incomplete clinical details or coding errors. Although validation studies in the UK Biobank have demonstrated high positive predictive values for major CVD outcomes (38,39), milder, nonhospitalized cases might not be captured, potentially leading to an underestimation of incident CVD events. Third, the potential bidirectional relationship between pulmonary dysfunction and CVD warrants thorough consideration (40). Despite attempts to control for prevalent lung diseases and mitigate reverse causality (e.g., 2-year blanking analysis), the possibility of reverse causality cannot be entirely ruled out. Furthermore, the observational design of the study prevents our findings from being used to infer causal relationships. Future studies are needed to disentangle this complex relationship. Fourth, despite adjusting for a wide range of confounders, residual confounding from unmeasured factors cannot be entirely ruled out. Finally, the study population consisted entirely of individuals of European ancestry and White ethnicity, highlighting the lack of racial and ethnic diversity and limiting the generalizability of the findings to other racial and ethnic groups. Additionally, the UK Biobank cohort is characterized by a healthy volunteer bias. These factors restrict the applicability of our findings and underscore the need for further research in more diverse and representative populations. By acknowledging these limitations, future research should prioritize inclusive cohorts with greater population diversity, incorporate dynamic assessments of lung function over time, and use innovative methodologies to further clarify the complex interplay between lung impairment and CVD, which could ultimately enhance the clinical relevance and applicability of our findings.

In summary, our findings indicate a significant link between lung function impairment and an increased risk of CVD incidence and mortality in individuals with T2D. This study underscores the potential of pulmonary function as a pivotal predictor for developing targeted prevention strategies against CVD in patients with T2D.

Acknowledgments. The authors thank the participants and staff of the UK Biobank for their dedication and contribution to the research.

Funding. This work was supported by Climbing Plan of Guangdong Provincial People’s Hospital grant DFJH2020022 and Guangdong Basic and Applied Basic Research Foundation-Provincial Enterprise Joint Fund grant 2022A1515220113.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.C. completed the initial data preparation and statistical analyses. C.C., Z.H., L.L., and B.S. were involved with the conception of the study. C.C., Y.F., and Y.H. drafted the manuscript. Y.F. and Y.H. provided the clinical expertise. All authors critically reviewed and contributed to the intellectual content of the manuscript and approved the final version of the manuscript. C.C. and Y.H. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Vanita R. Aroda.